Invitation code skyrocketed to over ten thousand yuan, China's AI new star exploded in popularity overnight

Author | Chai Xuchen

Editor | Zhou Zhiyu

Without any warm-up or spoilers, the Chinese AI team dropped a bombshell late at night, igniting another frenzy after DeepSeek.

In the early hours of March 6, the tech circle was flooded with news about an AI product named Manus, and the demand for invitation codes skyrocketed to over ten thousand yuan.

For Manus, the official description is extremely enticing—while large models are still outputting suggestions in dialogue boxes, Manus can directly put ideas into practice through independent thinking and planning, calling various tools in a virtual environment, writing and executing code, browsing the web, operating applications, and delivering final results.

In the demonstration video, it showcased three scenarios: uploading a compressed file containing multiple resumes, where Manus automatically unzips, analyzes the resumes, and provides suggestions and rankings based on user prompts; organizing property information and prices based on user requirements; generating analysis reports on the stock prices of NVIDIA and Tesla.

It is worth mentioning that Manus has "packaged" these capabilities into a cloud-based virtual environment, capable of processing tasks asynchronously. Users only need to input their requests, and Manus will activate a cloud virtual machine to autonomously complete data scraping, cross-validation, code writing, and visualization deployment, all without human supervision.

This "digital intern" model is reconstructing the boundaries of productivity—just as the GAIA benchmark test shows, Manus has surpassed human junior analysts by 9.3 percentage points in solving complex problems accurately. This could lead to a complete productivity revolution in personal workflows.

An AI industry insider pointed out to Wall Street News that Manus may drive a surge in demand for computing power, benefiting related cloud services, computing infrastructure, and AI application companies. The capital market reacted quickly; on the morning of March 6, the A-share AI agent sector led the market, with over 20 stocks, including Kute Intelligent, hitting the daily limit; computing concept stocks continued to strengthen, with over 20 stocks, including Hainan Huatie, also hitting the daily limit.

Clearly, the market is eager for a second DeepSeek to prove China's leading position in the AI race.

However, the rapidly emerging Manus has left the public with a lot of mystery.

Currently, Manus is not publicly available and can only be used through invitation codes, which are hard to come by. On second-hand trading platforms, many users are seeking invitation codes, but there are very few sellers, with some even asking for prices as high as 50,000 to 60,000 yuan.

As for the principles behind Manus, the official has not released any detailed technical reports. Some tech experts have analyzed that it is based on the foundational capabilities of Claude combined with the planning capabilities of post-training on open-source models, while also incorporating various intelligent agents to complete tasks through set workflows.

Industry insiders have analyzed to Wall Street News that the Manus team does not independently develop large models but rather provides scenario packaging services based on existing model capabilities, and this "shelling" is one of the secrets to its success As an AI product layer startup, Manus founder Xiao Hong does not pursue self-developed large models but instead enhances comprehensive capabilities through dynamic calls to multiple models like GPT-4 and Claude 3.

This does not hinder Manus from reaffirming that the AI agent track is accelerating in popularity.

In December last year, when Google released the Gemini 2.0 series, it clearly stated that it was "designed for the agent era." Microsoft has successively launched 10 AI agents and supports users in building autonomous agents on the Copilot Studio platform. Domestic tech giants such as ByteDance, Baidu, Tencent, and Alibaba, as well as AI companies like Zhipu, are also actively laying out the AI agent track.

CITIC Securities analysts pointed out that as more and more large companies lay out AI agent products and ecosystems, 2025 is expected to be the year of explosive commercial use of AI agents.

When AI can proactively complete tasks like humans, perhaps as Musk predicted, the era of general AGI will arrive in 2026. Investment researchers noted that future competition will no longer be a confrontation between humans and AI, but a competition for the efficiency of different human-machine coalitions' Scaling Law.

In this transformation, the real winners will be those organizations that first achieve "human-machine Scaling Law alignment." They not only possess the execution power of AI but also master the meta-capability to continuously evolve human-machine systems. When humans learn to harness Scaling Law with an architect's mindset, they will find that AI is not a replacer but a "capability amplifier" that pushes human civilization to new heights.

The industry has once again elevated Manus to a pedestal. However, the Manus team has shown great restraint in response.

On the afternoon of March 6, Manus AI partner Zhang Tao posted on social media, stating, "What everyone currently sees of Manus is still an infant in swaddling clothes, far from the experience we want to deliver in the official version. There is still much room for improvement in areas such as model hallucinations, deliverable friendliness, and operational speed."

Xiao Hong also mentioned that AI agents are still in the early stages and have not yet reached execution in the physical world; secondly, the capabilities of large models are still developing, and everything remains unpredictable.

Perhaps Manus may not be the final AI agent, but it undoubtedly raises people's expectations for AI by an order of magnitude.

Looking back, in this game interwoven with AI companies and capital, technological iteration has always been leapfrogging, leaving no time for a slow approach. It is still difficult to predict where the future world and AI industry will head, but the ongoing "sleepless nights" in the tech circle may be accelerating the arrival of the AGI endgame scenario into reality